🚀 VectorForgeML

High-Performance Machine Learning Framework for R & C++

---

📌 Overview

VectorForgeML is a next-generation machine learning framework designed to bridge the gap between R's simplicity and C++'s raw performance. Built from scratch, it focuses on understanding the mathematical foundations of ML while delivering a scalable, production-ready systems architecture.

Unlike traditional R packages that simply wrap C libraries, VectorForgeML implements core algorithms using:

• Raw C++ Pointers for graph-based models (Decision Trees, Random Forests)
• BLAS & LAPACK for hardware-accelerated Linear Algebra
• OpenMP for multi-core parallelism
• Zero-Copy data exchange between R and C++

The goal is to provide a Research-Ready Engine that allows data scientists to build complex pipelines in milliseconds.

Research Paper

The official research paper describing VectorForgeML is available here:

VectorForgeML Paper

---

⚡ Key Features

• 🚀 Blazing Fast: Optimized C++ backend ensures models train in record time compared to standard R implementations.
• 🔗 Modular Design: Component-based architecture with full support for Pipelines (ColumnTransformer, StandardScaler, PCA, etc.).
• 🔧 Hardware Acceleration: Native integration with BLAS/LAPACK for matrix operations.
• 📉 Memory Efficient: Custom allocators and index-based views (no deep copies).
• 🛠 Zero Dependencies: Core algorithms are implemented without heavy external ML dependencies.

---

🎉 Officially on CRAN

VectorForgeML is officially published on CRAN and distributed globally.

Install from CRAN (Recommended)

install.packages("VectorForgeML")
library(VectorForgeML)

---

🧠 Algorithms & Notebooks

We have implemented a wide range of algorithms, verified with real-world datasets on Kaggle.

🔹 Supervised Learning (Regression & Classification)

Algorithm	Type	Description	Kaggle Demo
Linear Regression	Regression	OLS with BLAS/LAPACK optimization.	View Notebook
Logistic Regression	Classification	Gradient Descent with Sigmoid activation.	View Notebook
Ridge Regression	Regression	L2 Regularized Linear Regression using Cholesky.	View Notebook
Softmax Regression	Classification	Multi-class classification with Log-Sum-Exp trick.	View Notebook
Decision Tree	Reg/Class	Recursive partitioning with distinct C++ graph pointers.	View Notebook
Random Forest	Ensemble	Parallelized ensemble of decision trees.	View Notebook
KNN	Reg/Class	K-Nearest Neighbors with std::partial_sort optimization.	View Notebook

🔹 Unsupervised Learning

Algorithm	Type	Description	Kaggle Demo
K-Means Clustering	Clustering	Lloyd's algorithm with efficient centroid updates.	View Notebook
PCA	Dim. Reduction	Principal Component Analysis via SVD/Eigen Decomposition.	View Notebook

🔹 Utilities & Pipelines

Feature	Description	Kaggle Demo
Pipeline	Chain multiple steps (Preprocessing -> Model) into one object.	View Notebook
Preprocessing	StandardScaler, MinMaxScaler, LabelEncoder, OneHotEncoder.	Included in Pipeline Demo
Metrics	accuracy_score, r2_score, f1_score, confusion_matrix.	Included in all Demos

---

🚀 Quick Start Example

Here is how you can build a powerful pipeline in just a few lines of R:

install.packages("VectorForgeML")
library(VectorForgeML)

df <- read.csv(system.file("dataset","winequality.csv", package="VectorForgeML"),sep=";")
y <- df$quality
X <- df; X$quality<-NULL

split <- train_test_split(X,y,0.2,42)

cat_cols <- names(X)[sapply(X,is.character)]
num_cols <- names(X)[!sapply(X,is.character)]

pre <- ColumnTransformer$new(
  num_cols,cat_cols,
  StandardScaler$new(),
  OneHotEncoder$new()
)

pipe <- Pipeline$new(list(
  pre,
  RandomForest$new(ntrees=100,max_depth=7,4,mode="classification")
))

pipe$fit(split$X_train,split$y_train)
pred <- pipe$predict(split$X_test)

cat("Accuracy:",accuracy_score(split$y_test,round(pred)),"\n")

---

🏗 Project Structure

VectorForgeML/
├── R/                  # R interface & utility functions
├── src/                # High-performance C++ backend
│   ├── LinearRegression.cpp
│   ├── DecisionTree.cpp
│   ├── RandomForest.cpp
│   └── ...
├── include/            # C++ Header files
├── public/             # Documentation Website (VectorForgeML.com)
├── scripts/            # Build & Doc generation scripts
└── README.md           # Project Documentation

---

👨‍� Developer

Mohd Musheer
Lead Developer & Architect

Passionate about High-Performance Computing (HPC) and System Design. VectorForgeML is a testament to building systems from first principles—combining rigorous mathematics with software engineering excellence.

---

📖 Citation

If you use VectorForgeML in research, please cite:

Musheer, M. (2026). VectorForgeML: High-Performance Machine Learning Framework. CRAN.
DOI: 10.32614/CRAN.package.VectorForgeML

Hackathon Submission

This project was submitted to the Quantum Sprint Hackathon 2026 on Devpost.

VectorForgeML is a high-performance machine learning framework with a C++ backend designed to accelerate model training in R.

Resources:

• Documentation: https://documentation.work.gd
• CRAN Package: https://cran.r-project.org/package=VectorForgeML
• Research Paper: paper/VectorForgeML_Paper.pdf

Hackathon

Submitted to: Quantum Sprint Hackathon 2026

📜 License

This project is licensed under the Apache License 2.0.

---

Made with ❤️ by Mohd Musheer